Statistical Sciences & Operations Research Seminar Virginia Commonwealth University

TITLE: Learning from streaming and imbalanced data SPEAKER: Dr. Bartosz Krawczyk Computer Science, VCU TIME & PLACE: Friday, Feb. 3, 2017, 11:00-12:00 Room 4119, Grace Harris Hall

ABSTRACT: Developing efficient classifiers, which are able to cope with big imbalanced and streaming data, especially with the presence of the so-called concept drift is currently one of the primary directions among the machine learning community. This presentation will highlight the most important issues these fields such as taking into account data characteristics, multi-class skewed distributions, adaptation to changes, and limited access to true class labels or novelty detection. In learning from imbalanced data, one should focus on analyzing not only the disproportion between classes, but also on other difficulties embedded in the nature of data. In mining data streams two challenges need to be addressed – how to detect a change and how to create an adaptive learning algorithm characterized by a low computational requirements. Additionally, a difficult scenario of drifting data stream with novel class appearance and evolving imbalance ratio will be discussed. This talk is going to be given from an engineering perspective and vital areas that would benefit from a more rigorous mathematical approach will be discussed.

Department of Statistical Sciences and Operations Research www.stat.vcu.edu (804) 828-0001